The company makes industrial products at its factory in Malaysia. In 2025 its Executive Director wanted staff to move from Microsoft Excel to Google Sheets, and to start using AI in daily work and on the factory floor.
Most of the team had worked the same way for years and were new to Google’s tools. Many records still started on paper.
Key takeaways
- Start with the records that cause the most double entry.
- Train every level together, so supervisors and directors share one way of working.
- Expect a runway of one to two months before results show, and plan a review.
- For work with your own data, pick tools that answer from that data.
The challenge
The primary challenge faced is manual data entry from hardcopy to softcopy, which is time-consuming and error-prone.”HR team, the client
The company wanted less time spent on tasks, faster reporting and fewer errors. The Executive Director also asked a harder question: how to bring AI into manufacturing with limited resources, and how to judge the return on the training.
- Six departments: logistics, purchasing and warehouse, accounts, HR, quality control and production.
- Data keyed in more than once, in sheets that did not link up.
The solution
Two days, in Malay
Delivered on site in Johor over two days in January 2026, with the Kuala Lumpur team joining online.
Day 1: Google Sheets and AI
One master spreadsheet for reporting, linked sheets so data is entered once and safe, effective prompting.
Day 2: shared dashboards and bringing colleagues along
So every department sees the same numbers.
A review with the leadership team
On 4 February 2026 we reviewed progress with the Executive Director and department heads.
A follow-up on NotebookLM
A half-day virtual session on 22 June 2026 showed staff and interns how to research and work with their data in NotebookLM.
The results
Two more departments are next.
| Measure | Before | After |
|---|---|---|
| Forms, first two departments | Paper only | Digital, alongside paper |
| Forms, next two departments | Paper only | In progress |
Source: Mereka AI Lab project records: leadership review, 4 February 2026, and follow-up session, 22 June 2026
At the February review the team said it needed a month or two before results would show. Staff time for data entry and a shortage of laptops held things back.
For cleaning data, it chose NotebookLM and Looker Studio over a general chatbot, to keep answers tied to its own records and avoid made-up results.
Lessons and what’s next
The company came back in June 2026 for the NotebookLM session. Next is finishing the remaining departments and measuring the return the Executive Director asked about: time saved on data entry in each department.
Sources and methods
The results come from our project records: the progress review with the Executive Director and department heads on 4 February 2026, and what the team reported at the NotebookLM follow-up on 22 June 2026. Time saved has not been measured yet.
